Despite increasing interest in the circumstances and outcomes of only children and increasing family complexity, the conceptualisation of only children has received limited scholarly attention. We raise issues involved in defining and identifying only children in social survey data, reflecting on the decisions researchers need to take. Illustrating the discussion with descriptive analyses of four British large-scale birth cohorts, we show it is possible to identify groups of individuals who correspond to different definitions of only children. Prevalence estimates obtained from the survey data show a similar trend over time as estimates from official fertility statistics. Researchers need to consider the choice of the most appropriate measure for a given research question, but data currently collected in many surveys limit how accurately any indicator of only children can reflect the chosen definition.